In the present project we designed and implemented a distance estimation system for objects. The system recognizes an object and marks its position in the pictures that were taken by the cameras.
Abstract
In the present project we designed and implemented a distance estimation system for objects. The system recognizes an object and marks its position in the pictures that were taken by the cameras. Each camera “sees” only a two-dimensional picture, and the estimation given is based on the angels from each camera to the object.
The algorithm
The implementation is executed in six phases:
A) Calibration of the system – the system is quite sensitive to it’s initial condition, and in order to get reasonable results we have to calibrate the camera’s position
B) Sampling pictures from two video cameras. Using an application for video signal sampling, we take pictures sampled from the video camera. We operate it once for each of the two cameras simultaneously
C) The user chooses an object, and marks it in the upper picture by a red square
D) Image possessing – Object recognition. Using least square error method, the program identifies the object, which was chosen and marks it by Blue Square
E) Using a lookup table, we find the angles of the object from each of the cameras. For each of the cameras, we get two angles: one in a vertical and the other in a horizontal plane
F) Mathematical calculation of the distance, based on the angles we found in the previous step, and the distance between the cameras – which was given to the system in the first (initiation) step
Results
The system gives an accurate estimation almost in real-time, giving option for other applications to use this information.
here is an example of the accuracy of the system:
as we can view in the graph, the maximal estimation error the system produced in 5 meters was 4 cm (that is, for an object standing 4.5 meters away from the system, the system believes that the objects stands 4.46 meters – an error of 0.88 percent)
the system does not seems to be sensitive to the distance of the object or to its relative angle.
Acknowledgments
We would like to thank our supervisor Johanan Erez for his support and guidance throughout this project.
we would also like to thank the Ollendorff Minerva Center Fund which supported this project.